Neural Network-Based Intelligent Control of Continuous Flow Ohmic Heating Systems for Enhanced Dynamic Performance and Sustainable Food Processing

Abstract

Continuous flow Ohmic heating (CFOH) is a sustainable thermal processing technology that enables rapid volumetric heating through the electrical resistance of food materials. However, the strong nonlinear coupling between electrical conductivity, temperature, and heat transfer dynamics complicates accurate temperature regulation and stable process operation. This study proposes and evaluates advanced neural network (NN)-based control strategies for nonlinear CFOH systems using nonlinear autoregressive moving average level-2 (NARMA-L2) and model reference control (MRC) architectures. A real-time validated pilot-scale CFOH model implemented in MATLAB/Simulink was utilised to develop, train, and evaluate the controllers under realistic food processing conditions using sweet and sour sauce as the working fluid. The proposed framework integrates dynamic performance analysis, robustness evaluation, energy efficiency assessment, and indirect greenhouse gas (GHG) emission analysis within an integrated evaluation platform. Controller robustness was evaluated under variations in electrical conductivity, flow rate, inlet temperature, sensor noise, and setpoint disturbances. Within the validated simulation framework, the results demonstrate that the NARMA-L2 controller achieved faster dynamic response, reduced settling time, improved stability, zero overshoot, and lower steady-state energy consumption compared to other evaluated strategies. The NN-based controllers also maintained stable performance under varying operating conditions, demonstrating improved adaptability to nonlinear process behaviour. Overall, the proposed NN-based controllers demonstrate strong potential for enhancing process efficiency, operational stability, and sustainability in industrial CFOH applications.